[R] Issues installing package: Is tensorflow <2 required?

[jansim]

I wanted to install & setup the package on my machine in order to work on #329 and ran into issues with resolving dependencies on a Mac with M1 chip.

I first tried installing the package using a conda env with python 3.7, however, this failed due to python 3.7 only being available for x86_64 architectures and modern versions of R running under arm64e. While python 3.7 can be be run with rosetta emulation it seems that the interaction with R is not possible across architectures.

My next step was to try using conda envs with python 3.8 and 3.9 to install the package, however, this failed due to tensorflow 1.15 not being available for either of these two python versions. Since both versions are officially supported by AIF360, this left me quite surprised as the requirements.txt does specify tensorflow>=1.13.1,<2 which afaik cannot be fulfilled with either python 3.8 or 3.9? FWIW, it seems that the CI is just ignoring the <2 requirement and will happily install tensorflow 2.x, so maybe the requirement can be dropped? 🤔